williamp/osmocom-analog
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

C-Netz: Calibration/fix processing of tone levels on TX and RX side

Browse Source 
FSK and compander levels are adjusted.

Eliminate offsets between subsequent speech chunks. This is done by
high-pass filter. An offset is not passed to the filter.

Do correct audio processing chain:
 time compress -> compressor -> scrambler / pre-emphasis -> TX
 RX -> de-scrambler / de-emphasis -> expander -> time expand
This commit is contained in:
Andreas Eversberg
2016-05-13 08:56:26 +02:00
parent 5a9c9c7401
commit 46fa72894e
4 changed files with 83 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files

88
src/cnetz/dsp.c
View File

@@ -25,23 +25,26 @@
#include <errno.h>
#include "../common/debug.h"
#include "../common/timer.h"
#include "../common/call.h"
#include "cnetz.h"
#include "sysinfo.h"
#include "telegramm.h"
#include "dsp.h"
/* test function to mirror received audio from ratio back to radio */
//#define TEST_SCRABLE
/* test the audio quality after cascading two scramblers (TEST_SCRABLE must be defined) */
//#define TEST_UNSCRABLE
//#define TEST_SCRAMBLE
/* test the audio quality after cascading two scramblers (TEST_SCRAMBLE must be defined) */
//#define TEST_UNSCRAMBLE
#define PI M_PI
#define FSK_DEVIATION 16384
#define FSK_DEVIATION 10000
#define COMPANDER_0DB 20000
#define BITRATE 5280.0 /* bits per second */
#define BLOCK_BITS 198 /* duration of one time slot including pause at beginning and end */
#define CUT_OFF_OFFSET 300.0 /* cut off frequency for offset filter (level correction between subsequent audio chunks) */
#ifdef TEST_SCRABLE
#ifdef TEST_SCRAMBLE
jitter_t scrambler_test_jb;
scrambler_t scrambler_test_scrambler1;
scrambler_t scrambler_test_scrambler2;
@@ -77,6 +80,7 @@ int dsp_init_sender(cnetz_t *cnetz, int measure_speed, double clock_speed[2], do
{
int rc = 0;
double size;
double RC, dt;
PDEBUG(DDSP, DEBUG_DEBUG, "Init FSK for 'Sender'.\n");
@@ -136,9 +140,14 @@ int dsp_init_sender(cnetz_t *cnetz, int measure_speed, double clock_speed[2], do
/* init compander, according to C-Netz specs, attack and recovery time
* shall not exceed according to ITU G.162 */
init_compander(&cnetz->cstate, 8000, 5.0, 22.5, 32767);
init_compander(&cnetz->cstate, 8000, 5.0, 22.5, COMPANDER_0DB);
#ifdef TEST_SCRABLE
/* use this filter to compensate level changes between two subsequent audio chunks */
RC = 1.0 / (CUT_OFF_OFFSET * 2.0 *3.14);
dt = 1.0 / cnetz->sender.samplerate;
cnetz->offset_factor = RC / (RC + dt);
#ifdef TEST_SCRAMBLE
rc = jitter_create(&scrambler_test_jb, cnetz->sender.samplerate / 5);
if (rc < 0) {
PDEBUG(DDSP, DEBUG_ERROR, "Failed to init jitter buffer for scrambler test!\n");
@@ -565,8 +574,8 @@ void sender_receive(sender_t *sender, int16_t *samples, int length)
/* measure rx sample speed */
calc_clock_speed(cnetz, length, 0, 0);
#ifdef TEST_SCRABLE
#ifdef TEST_UNSCRABLE
#ifdef TEST_SCRAMBLE
#ifdef TEST_UNSCRAMBLE
scrambler(&scrambler_test_scrambler1, samples, length);
#endif
jitter_save(&scrambler_test_jb, samples, length);
@@ -711,12 +720,15 @@ again:
if (*spl == -32768) {
/* marker found to insert new chunk of audio */
jitter_load(&cnetz->sender.audio, speech_buffer, 100);
/* 1. compress dynamics */
compress_audio(&cnetz->cstate, speech_buffer, 100);
/* 2. upsample */
speech_length = samplerate_upsample(&cnetz->sender.srstate, speech_buffer, 100, speech_buffer);
/* 3. scramble */
if (cnetz->scrambler)
scrambler(&cnetz->scrambler_tx, speech_buffer, speech_length);
/* pre-emphasis is done by cnetz code, not by common code */
/* pre-emphasis makes bad sound in conjunction with scrambler, so we disable */
/* 4. pre-emphasis is done by cnetz code, not by common code */
/* pre-emphasis is only used when scrambler is off, see FTZ 171 TR 60 Clause 4 */
if (cnetz->pre_emphasis && !cnetz->scrambler)
pre_emphasis(&cnetz->estate, speech_buffer, speech_length);
speech_pos = 0;
@@ -751,7 +763,7 @@ void sender_send(sender_t *sender, int16_t *samples, int length)
/* measure tx sample speed */
calc_clock_speed(cnetz, length, 1, 0);
#ifdef TEST_SCRABLE
#ifdef TEST_SCRAMBLE
jitter_load(&scrambler_test_jb, samples, length);
scrambler(&scrambler_test_scrambler2, samples, length);
return;
@@ -775,3 +787,55 @@ void sender_send(sender_t *sender, int16_t *samples, int length)
#endif
}
/* unshrink audio segment from the duration of 60 bits to 12.5 ms */
void unshrink_speech(cnetz_t *cnetz, int16_t *speech_buffer, int count)
{
int16_t *spl;
int32_t value;
int pos, i;
double x, y, x_last, y_last, factor;
/* fix offset between speech blocks by using high pass filter */
/* use first sample as previous sample, so we don't have a level jump between two subsequent audio chunks */
x_last = speech_buffer[0];
y_last = cnetz->offset_y_last;
factor = cnetz->offset_factor;
for (i = 0; i < count; i++) {
x = (double)speech_buffer[i];
/* high-pass to remove low level frequencies, caused by level jump between audio chunks */
y = factor * (y_last + x - x_last);
x_last = x;
y_last = y;
value = (int32_t)y;
if (value < -32768.0)
value = -32768.0;
else if (value > 32767)
value = 32767;
speech_buffer[i] = value;
}
cnetz->offset_y_last = y_last;
/* 4. de-emphasis is done by cnetz code, not by common code */
/* de-emphasis is only used when scrambler is off, see FTZ 171 TR 60 Clause 4 */
if (cnetz->de_emphasis && !cnetz->scrambler)
de_emphasis(&cnetz->estate, speech_buffer, count);
/* 3. descramble */
if (cnetz->scrambler)
scrambler(&cnetz->scrambler_rx, speech_buffer, count);
/* 2. decompress time */
count = samplerate_downsample(&cnetz->sender.srstate, speech_buffer, count, speech_buffer);
/* 1. expand dynamics */
expand_audio(&cnetz->cstate, speech_buffer, count);
/* to call control */
spl = cnetz->sender.rxbuf;
pos = cnetz->sender.rxbuf_pos;
for (i = 0; i < count; i++) {
spl[pos++] = speech_buffer[i];
if (pos == 160) {
call_tx_audio(cnetz->sender.callref, spl, 160);
pos = 0;
}
}
cnetz->sender.rxbuf_pos = pos;
}